rougier
commented
9 years ago I plainly agree with you. Reproducibility is definitely a more and more prominent problem in computational science and any published papers with data and code should be made publicly available on sites like github. Ideally, we should even be able to run the code directly in the browser (e.g. via an online virtual machine for example). But I'm surprised teams you've contacted were not able to send you data and sources, 3 years is not that long, even in digital era.
As a matter of fact, I'm currently suffering the same problem with a model from neuroscience. Authors were kindly enough to send me the sources (they were not initially published) but then I had to deal with 6000 lines of Delphi... No ways to compile it but the binary was working ok. In the end, it took us around 3 months to recode the model. Once done, we immediately put it on github (https://github.com/rougier/Neurosciences).
But I would prefer an academic solution instead of github, because it might disappear in the future.
Thank you for putting your scripts in a public repository. This is a much needed positive example among academics. In the last six months I have contacted three different teams who each published a paper. I asked them if they still had the data sets and scripts for doing a particular analysis (analyzing the efficiency of a keyboard layout). They did not have the scripts or the data sets (3-5 years after publication). The only thing left of their work is the published paper. As a researcher I would like to use their script and test other languages with the same script to provide some sort of comparative feedback on the methods they used. Not possible due to "erosion of the base". Thank you for presenting a positive example. Let's hope link rot does not reach to github.